Fall recognition analysis and alarm method based on smart watch

Abstract

The invention relates to the field of intelligent wearable equipment, in particular to a fall identification analysis and alarm method based on an intelligent watch. Comprising the following steps: acquiring acceleration data and angular velocity data, and filtering the acceleration data and the angular velocity data to obtain filtered acceleration data and angular velocity data; dynamic reverse weighting features are extracted and obtained; on the basis of the dynamic reverse weighting features and the filtered acceleration data and angular velocity data, calculating a tumble judgment value, and preliminarily judging whether a tumble event occurs or not; a composite weighted judgment mechanism is introduced, and a weighted judgment value is calculated; and calculating alarm delay response time based on the weighted decision value, comparing the alarm delay response time with an alarm threshold value, judging whether a fall event occurs or not, and triggering an alarm after confirmation. The technical problem that an effective noise filtering mechanism is lacked in processing of acceleration data and angular velocity data, so that noise in sensor data interferes with falling judgment, and the accuracy of a judgment result is affected is solved.

Description

Technical Field

[0001] The present invention relates to the field of smart wearable devices, and in particular to a fall recognition analysis and alarm method based on a smart watch. Background Art

[0002] With the rapid development of smart wearable device technology, smart watches have become an important health monitoring tool in people's daily lives. At present, many smart watches can collect users' motion data in real time through built-in sensors such as accelerometers and gyroscopes, and perform fall detection and health management. In order to improve the accuracy of fall detection, researchers have proposed a variety of analysis methods based on acceleration and angular velocity data, including simple algorithms based on threshold judgment and intelligent recognition technology based on machine learning. By collecting and analyzing users' motion data, existing technologies have made certain progress in the sensitivity, real-time and reliability of fall detection. In order to further improve the accuracy of fall recognition, related technologies have also begun to integrate more sensor data and algorithm optimization, such as combining filtering technology to reduce noise, using dynamic features to identify the changing laws of acceleration and angular velocity, and using intelligent algorithms for multi-dimensional judgment, striving to provide more accurate fall alarm services.

[0003] However, the existing fall recognition analysis and alarm methods have the following technical problems: there is a lack of effective noise filtering mechanism in the processing of acceleration data and angular velocity data, which causes the noise in the sensor data to interfere with the fall judgment, thereby affecting the accuracy of the judgment result; the traditional method fails to fully utilize the relationship between the dynamic changes of acceleration data and angular velocity data, ignores the timing characteristics and fluctuation amplitude of the user's motion data, makes the sensitivity of fall detection low, and cannot identify sudden fall events in time; in addition, most of the existing fall recognition methods have a delay in the alarm response time and cannot accurately trigger the alarm in a short time, which brings potential safety risks to users. Summary of the invention

[0004] The present invention provides a fall recognition analysis and alarm method based on a smart watch to solve the problem that there is a lack of an effective noise filtering mechanism in the processing of acceleration data and angular velocity data, which causes the noise in the sensor data to interfere with the fall judgment, thereby affecting the accuracy of the judgment result; the traditional method fails to fully utilize the relationship between the dynamic changes of acceleration data and angular velocity data, ignores the timing characteristics and fluctuation amplitude of the user's motion data, makes the sensitivity of fall detection low, and cannot timely identify sudden fall events; in addition, most existing fall recognition methods have a delay in alarm response time, and cannot accurately trigger an alarm in a short time, which brings potential safety risks to users.

[0005] The fall recognition analysis and alarm method based on a smart watch of the present invention specifically includes the following technical solutions:

[0006] The fall recognition analysis and alarm method based on the smart watch includes the following steps:

[0007] S1: Acquire acceleration data and angular velocity data, filter the acceleration data and angular velocity data to obtain filtered acceleration data and angular velocity data; and extract dynamic inverse weighted features from the filtered acceleration data and angular velocity data;

[0008] S2: Based on the dynamic reverse weighted features and the filtered acceleration data and angular velocity data, the fall judgment value is calculated to preliminarily determine whether a fall event has occurred; and a composite weighted judgment mechanism is introduced to calculate the weighted judgment value; based on the weighted judgment value, the alarm delay response time is calculated and compared with the alarm threshold to determine whether it is a fall event, and the alarm is triggered after confirmation.

[0009] Preferably, the S1 specifically includes:

[0010] The calculation formulas for filtered acceleration data and angular velocity data are as follows:

[0011]

[0012] in, is the filtered acceleration data and angular velocity data; t is the time index variable; N is the filter window size; i is the time offset index; σ is the standard deviation of the Gaussian function; exp is the exponential function operation; x(ti) is the input data at time ti; is the time rate of change of the input data, which serves as a differential term; γ is the weight coefficient of the differential term.

[0013] Preferably, the S1 specifically includes:

[0014] The dynamic inverse weighted feature is calculated by the relationship between the changes in the filtered acceleration data and the angular velocity data, and evaluates the movement changes associated with the fall at the current moment.

[0015] Preferably, the S1 specifically includes:

[0016] The dynamic inverse weighted feature extraction formula is as follows:

[0017]

[0018] Among them, P(t) is the dynamic inverse weighted feature at time t; t is the time index variable; 3 is a constant, indicating the x, y, and z axes involved in the calculation; j is the axis index variable; is the change in the filtered acceleration data; is the component of the filtered acceleration data on the jth axis at time t; λ is the inverse weighting factor; sign is the sign function; is the change in the filtered angular velocity data; It is the component of the filtered angular velocity data on the ,th axis at time t.

[0019] Preferably, the S2 specifically includes:

[0020] The composite weighted judgment mechanism calculates a weighted judgment value based on the dynamic reverse weighted feature. The weighted judgment value measures the possibility of a fall event by calculating the difference between the dynamic reverse weighted feature at the current moment and the dynamic reverse weighted feature at the past moment.

[0021] Preferably, the S2 specifically includes:

[0022] The weighted judgment value formula is as follows:

[0023]

[0024] Where W(t) is the weighted decision value; t is the time index variable; is the time window size; ζ is the time index variable starting from time t; P(t) and P(t-ζ) are the dynamic inverse weighted features at time t and time t-ζ respectively; exp is the exponential function operation; is the standard deviation of the exponential function.

[0025] Preferably, the S2 specifically includes:

[0026] The alarm delay effect time is calculated according to the weighted judgment value and the dynamic reverse weighted feature. The alarm delay effect time is obtained by introducing a weighted integral term and a Gaussian function and combining the weighted judgment value and the dynamic reverse weighted feature. The specific implementation formula is:

[0027]

[0028] Among them, Δt is the alarm delay response time; is the weighted integral term; is a Gaussian function; t 0 Indicates the current moment; is the standard deviation of the Gaussian function; max(P(t)) is the In, the maximum value of the dynamic inverse weighted feature P(t); is the component of the filtered acceleration data on the jth axis at time t; It is the component of the filtered angular velocity data on the jth axis at time t.

[0029] Preferably, the S2 specifically includes:

[0030] Set the alarm threshold and compare the calculated alarm delay response time with the alarm threshold. When the alarm delay response time is greater than the alarm threshold, it is considered not a fall event and the alarm is not triggered. When the alarm delay response time is less than or equal to the alarm threshold, it is determined to be a fall event and an alarm is triggered according to the alarm delay response time to remind the user to take measures.

[0031] The beneficial effects of the technical solution of the present invention are:

[0032] 1. The present invention adopts an improved Gaussian filtering algorithm to filter the collected acceleration data and angular velocity data, effectively removes noise, improves the quality of acceleration data and angular velocity data, and can obtain more accurate sensor data, providing reliable input for subsequent dynamic inverse weighted feature extraction, thereby improving the accuracy and stability of fall recognition.

[0033] 2. The present invention can comprehensively consider the amplitude of acceleration and angular velocity changes by extracting dynamic inverse weighted features from the filtered acceleration data and angular velocity data. The calculation method of the dynamic inverse weighted features reflects the fluctuation amplitude of the filtered acceleration data and angular velocity data, effectively enhancing the sensitivity of fall events, especially in complex and atypical fall scenarios, and can more accurately capture the dynamic changes of falls, thereby improving the sensitivity and accuracy of fall judgment.

[0034] 3. The present invention not only improves the recognition accuracy of fall events, but also optimizes the alarm delay response by introducing a composite weighted judgment mechanism. The composite weighted judgment mechanism combines the difference in dynamic inverse weighted characteristics between the current moment and the historical moment, and calculates the alarm delay effect time through weighted integration and Gaussian function, ensuring that the alarm can be triggered quickly after the fall occurs, which can improve the response speed and reliability in complex environments, effectively reduce misjudgments and missed judgments, and timely ensure user safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of the fall recognition analysis and alarm method based on a smart watch described in the present invention. DETAILED DESCRIPTION

[0036] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0038] The specific scheme of the fall recognition analysis and alarm method based on the smart watch provided by the present invention is described in detail below with reference to the accompanying drawings.

[0039] Refer to the attached Figure 1 , which shows a flow chart of a fall recognition analysis and alarm method based on a smart watch provided by an embodiment of the present invention, the method comprising the following steps:

[0040] S1: Acquire acceleration data and angular velocity data, filter the acceleration data and angular velocity data to obtain filtered acceleration data and angular velocity data; and extract dynamic inverse weighted features from the filtered acceleration data and angular velocity data;

[0041] First, the smart watch collects the user's motion data through the built-in acceleration sensor and angular velocity sensor. The motion data includes acceleration data a(t)=(a 1 (t), a 2 (t), a 3 (t)) and angular velocity data ω(t) = (ω 1 (t),ω 2 (t),ω 3 (t)), where t is the time index variable, a(t) is the acceleration data at time t, and a 1 (t), a 2 (t), a 3 (t) are the components of the acceleration data at time t on the x-axis, y-axis, and z-axis; ω(t) is the angular velocity data at time t, ω 1 (t),ω 2 (t),ω 3 (t) are the components of the angular velocity data on the x-axis, y-axis, and z-axis respectively.

[0042] In order to remove noise, the improved Gaussian filtering algorithm is used to filter the collected acceleration data and angular velocity data to obtain filtered acceleration data and angular velocity data. The filtering formula is as follows:

[0043]

[0044] in, is the filtered acceleration data and angular velocity data, specifically the component of the filtered acceleration data on the x-axis at time t Component on the y-axis The component on the z-axis and the component of the filtered angular velocity data on the x-axis at time t Component on the y-axis The component on the z-axis t is the time index variable; N is the filter window size, which is a positive integer and determines the range of acceleration data and angular velocity data considered by the filter formula when smoothing, so as to make the filtering effect smoother. The filter window size is set according to the specific implementation scenario; i is the time offset index, which is used to represent a certain historical or future moment relative to the moment t; σ is the standard deviation of the Gaussian function, which is used to control the width of the Gaussian distribution and thus the degree of smoothing, and is set according to the expert experience method; exp is the exponential function operation; x(ti) is the input data at the moment ti, and the input data is specifically the components of the acceleration data and angular velocity data on the x-axis, y-axis, and z-axis; is the time rate of change of the input data, which is used as the differential term; γ is the weight coefficient of the differential term, which is set according to the expert experience method.

[0045] The filtering formula combines the input data and the time rate of change of the input data, which can not only smooth the input data but also respond more sensitively to rapid changes in acceleration and angular velocity.

[0046] Furthermore, dynamic inverse weighted features are extracted from the filtered acceleration data and angular velocity data. The dynamic inverse weighted features are calculated through the relationship between the changes in the filtered acceleration data and the angular velocity data, reflecting the fluctuation amplitude of the filtered acceleration data and the angular velocity data, and are used to evaluate the motion changes related to the fall at the current moment, effectively enhancing the reliability and response speed of the fall judgment.

[0047] The dynamic inverse weighted feature extraction formula is as follows:

[0048]

[0049] Among them, P(t) is the dynamic inverse weighted feature at time t, which is calculated by the change of the filtered acceleration data and angular velocity data, reflecting the fluctuation amplitude of the filtered acceleration data and angular velocity data at time t; t is the time index variable; 3 is a constant, indicating the x, y, and z axes involved in the calculation; j is the axis index variable, which is used to traverse each axis during the calculation process, and the constants 1, 2, and 3 correspond to the x, y, and z axes respectively; is the change in the filtered acceleration data, indicating the change in the filtered acceleration data at the jth axis index at time t. And the filtered acceleration data at time t-1 The difference is calculated; is the component of the filtered acceleration data on the jth axis at time t; is the component of the filtered acceleration data on the jth axis at time t-1; λ is the inverse weighting factor, which is used to adjust the weight ratio between the changes in the filtered acceleration data and the angular velocity data, and determines the degree of influence of the changes in the filtered acceleration data and the angular velocity data on the feature extraction process. The value range is between 0.1 and 10, and the specific value is set by expert experience; sign is a sign function, which is used to determine the sign of the product of the changes in the filtered acceleration data and the angular velocity data, reflecting the relationship between the directions of change of acceleration and angular velocity. If the product of the changes in the filtered acceleration data and the angular velocity data is positive, 1 is returned; if the product of the changes in the filtered acceleration data and the angular velocity data is negative, -1 is returned; if the product of the changes in the filtered acceleration data and the angular velocity data is zero, 0 is returned; is the change in the filtered angular velocity data, indicating the change in the filtered angular velocity data at the jth axis index at time t. And the filtered angular velocity data at time t-1 The difference is calculated; is the component of the filtered angular velocity data on the jth axis at time t; It is the component of the filtered angular velocity data on the jth axis at time t-1.

[0050] The dynamic inverse weighted feature extraction formula comprehensively considers the changes in the filtered acceleration data and angular velocity data and their mutual relationship, and dynamically weights the effects of changes in acceleration and angular velocity, thereby improving the sensitivity and accuracy of fall detection, especially in complex and atypical fall scenarios.

[0051] S2: Based on the dynamic reverse weighted features and the filtered acceleration data and angular velocity data, the fall judgment value is calculated to preliminarily determine whether a fall event has occurred; and a composite weighted judgment mechanism is introduced to calculate the weighted judgment value; based on the weighted judgment value, the alarm delay response time is calculated and compared with the alarm threshold to determine whether it is a fall event, and the alarm is triggered after confirmation.

[0052] Based on the dynamic inverse weighted features, fall recognition is further performed, and a preliminary judgment is made as to whether a fall event has occurred by calculating the fall judgment value.

[0053] The fall determination value formula is as follows:

[0054]

[0055] Among them, F(t) is the fall judgment value, which reflects the fluctuation amplitude and change trend of the motion state; t is the time index variable; P(t) is the dynamic inverse weighted feature at time t; T is the reference value of the dynamic inverse weighted feature, which is set according to the expert experience method; ΔP(t) is the change of the dynamic inverse weighted feature, which is calculated by the difference between the dynamic inverse weighted feature P(t) at time t and the dynamic inverse weighted feature P(t-1) at time t-1; j is the axis index variable; is the component of the filtered acceleration data on the ,th axis at time t; It is the component of the filtered angular velocity data on the ,th axis at time t.

[0056] The fall determination threshold is set according to the specific implementation scenario. If the fall determination value is less than the fall determination threshold, it is determined that no fall event has occurred at the current moment. If the fall determination value is greater than or equal to the fall determination threshold, it is preliminarily considered that a fall event has occurred at the current moment.

[0057] After initially considering that a fall event has occurred at the current moment, in order to further improve the accuracy of fall identification, a composite weighted judgment mechanism is introduced. The composite weighted judgment mechanism comprehensively considers the difference between the dynamic reverse weighted features at the current moment and the dynamic reverse weighted features at the past moment, and combines an exponential function to enhance the accuracy and reliability of fall judgment. The specific process of the composite weighted judgment mechanism is as follows:

[0058] First, based on the dynamic reverse weighted features, a weighted judgment value is calculated. The weighted judgment value measures the possibility of a fall event by calculating the difference between the dynamic reverse weighted features at the current moment and the dynamic reverse weighted features at the past moment. The larger the value of the weighted judgment value, the more significant the difference between the dynamic reverse weighted features at the current moment and the dynamic reverse weighted features at the past moment, reflecting a greater movement change or fall risk, which means a higher probability of a fall.

[0059] The weighted judgment value formula is as follows:

[0060]

[0061] Where W(t) is the weighted judgment value, which is used to measure the difference between the dynamic inverse weighted features at time t and the dynamic inverse weighted features at time t-ζ; t is the time index variable; is the time window size, which indicates the maximum time step from time t to the past, and is set according to the specific implementation scenario; ζ is the time index variable from time t to the past, which indicates the time difference between time t and the past, and increases from 0 to the time window size Until; P(t) and P(t-ζ) are the dynamic inverse weighted features at time t and time t-ζ respectively; exp is the exponential function operation; It is the standard deviation of the exponential function, which is used to control the width of the exponential function. The smaller the standard deviation of the exponential function, the smaller the influence of the dynamic inverse weighted features farther away from the current moment on the calculation result. The standard deviation of the exponential function is set according to the expert experience method.

[0062] Furthermore, the alarm delay response time is calculated based on the weighted judgment value and the dynamic inverse weighted feature. The alarm delay response time is calculated by combining the weighted judgment value and the dynamic inverse weighted feature. A weighted integral term and a Gaussian function are introduced to measure the time difference between the occurrence of a fall and the triggering of the alarm, so as to ensure that the alarm can be accurately triggered in the shortest time.

[0063] The calculation formula for the alarm delay response time is as follows:

[0064]

[0065] Among them, Δt is the alarm delay response time, which means the time difference response between the fall occurrence and the alarm triggering after the fall event is detected; is the time window size; is the weighted integral term, calculated over the time window The weighted sum of the internal dynamic inverse weighted features, the dynamic inverse weighted features are weighted by Gaussian function, and the attenuation effect of the dynamic inverse weighted features within the historical time range is measured; P(t) is the dynamic inverse weighted feature at time t; is a Gaussian function used to calculate t 0 The dynamic inverse weighted features before time t are attenuated; 0 Indicates the current moment; is the standard deviation of the Gaussian function, which is set according to the expert experience method; W(t) is the weighted judgment value; max(P(t)) is the value in the time window The maximum value of the dynamic inverse weighted feature P(t).

[0066] Finally, the alarm threshold is set according to the specific implementation scenario, and the calculated alarm delay response time is compared with the alarm threshold. If the alarm delay response time is greater than the alarm threshold, it is considered not a fall event and the alarm is not triggered. If the alarm delay response time is less than or equal to the alarm threshold, it is determined to be a fall event, and the alarm is triggered according to the alarm delay response time to promptly remind the user or relevant personnel to take measures. The alarm method is set according to the specific implementation scenario. For example, when the smart watch detects a fall event and triggers the alarm, it reminds the wearer by vibrating and making a sound, and at the same time sends a text message or makes a call to the preset emergency contact to inform them of the fall event.

[0067] In summary, the fall recognition analysis and alarm method based on smart watches has been completed.

[0068] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0069] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A fall recognition analysis and alarm method based on a smart watch, characterized in that: The following steps are involved: S1: Acquire acceleration data and angular velocity data, filter the acceleration data and angular velocity data to obtain filtered acceleration data and angular velocity data; and extract dynamic inverse weighted features from the filtered acceleration data and angular velocity data; S2: Based on the dynamic reverse weighted features and the filtered acceleration data and angular velocity data, the fall judgment value is calculated to preliminarily determine whether a fall event has occurred; and a composite weighted judgment mechanism is introduced to calculate the weighted judgment value; based on the weighted judgment value, the alarm delay response time is calculated and compared with the alarm threshold to determine whether it is a fall event, and the alarm is triggered after confirmation.

2. The fall recognition, analysis and alarm method based on a smart watch according to claim 1, characterized in that: The S1 specifically includes: The calculation formulas for filtered acceleration data and angular velocity data are as follows: in, is the filtered acceleration data and angular velocity data; t is the time index variable; N is the filter window size; i is the time offset index; σ is the standard deviation of the Gaussian function; exp is the exponential function operation; x(ti) is the input data at time ti; is the time rate of change of the input data, which serves as a differential term; γ is the weight coefficient of the differential term.

3. The fall recognition, analysis and alarm method based on a smart watch according to claim 2, characterized in that: The S1 specifically includes: The dynamic inverse weighted feature is calculated by the relationship between the changes in the filtered acceleration data and the angular velocity data, and evaluates the movement changes associated with the fall at the current moment.

4. The fall recognition, analysis and alarm method based on a smart watch according to claim 3 is characterized in that: The S1 specifically includes: The dynamic inverse weighted feature extraction formula is as follows: Among them, P(t) is the dynamic inverse weighted feature at time t; t is the time index variable; 3 is a constant, indicating the x, y, and z axes involved in the calculation; j is the axis index variable; is the change in the filtered acceleration data; is the component of the filtered acceleration data on the jth axis at time t; λ is the inverse weighting factor; sign is the sign function; is the change in the filtered angular velocity data; It is the component of the filtered angular velocity data on the ,th axis at time t.

5. The fall recognition analysis and alarm method based on a smart watch according to claim 1, characterized in that: The S2 specifically includes: The composite weighted judgment mechanism calculates a weighted judgment value based on the dynamic reverse weighted feature. The weighted judgment value measures the possibility of a fall event by calculating the difference between the dynamic reverse weighted feature at the current moment and the dynamic reverse weighted feature at the past moment.

6. The fall recognition, analysis and alarm method based on a smart watch according to claim 5, characterized in that: The S2 specifically includes: The weighted judgment value formula is as follows: Where W(t) is the weighted decision value; t is the time index variable; is the time window size; ζ is the time index variable starting from time t; P(t) and P(t-ζ) are the dynamic inverse weighted features at time t and time t-ζ respectively; exp is the exponential function operation; is the standard deviation of the exponential function.

7. The fall recognition, analysis and alarm method based on a smart watch according to claim 6, characterized in that: The S2 specifically includes: The alarm delay effect time is calculated according to the weighted judgment value and the dynamic reverse weighted feature. The alarm delay effect time is obtained by introducing a weighted integral term and a Gaussian function and combining the weighted judgment value and the dynamic reverse weighted feature. The specific implementation formula is: Among them, Δt is the alarm delay response time; is the weighted integral term; is a Gaussian function; t0 represents the current moment; is the standard deviation of the Gaussian function; max(P(t)) is the In, the maximum value of the dynamic inverse weighted feature P(t); is the component of the filtered acceleration data on the jth axis at time t; It is the component of the filtered angular velocity data on the jth axis at time t.

8. The fall recognition, analysis and alarm method based on a smart watch according to claim 7, characterized in that: The S2 specifically includes: Set the alarm threshold and compare the calculated alarm delay response time with the alarm threshold. When the alarm delay response time is greater than the alarm threshold, it is considered not a fall event and the alarm is not triggered. When the alarm delay response time is less than or equal to the alarm threshold, it is determined to be a fall event and an alarm is triggered according to the alarm delay response time to remind the user to take measures.

Images